Chapter 11: Inference for Distributions of Categorical Data

The chi-square goodness-of-fit test evaluates whether observed counts of a single categorical variable align with theoretically expected frequencies, requiring students to calculate the test statistic by comparing observed and expected values, determine appropriate degrees of freedom, and interpret p-values from the chi-square distribution. A critical foundation involves understanding the necessary conditions for valid inference, including data obtained through random sampling, sufficiently large sample sizes, and a minimum expected count of five in each category to ensure the chi-square distribution provides accurate results. The chapter extends these concepts to the chi-square test for homogeneity, which compares how a categorical variable is distributed across multiple independent populations or experimental treatment groups, allowing researchers to determine whether distributions differ meaningfully. The chi-square test for independence addresses a related but distinct question about two categorical variables measured within a single population, testing whether these variables are statistically associated or independent of one another. Students learn to interpret standardized residuals, which reveal which specific categories contribute most substantially to the overall chi-square statistic, enabling deeper insight into where observed and expected values diverge most dramatically. Throughout the chapter, emphasis falls on recognizing that statistical significance merely indicates a meaningful deviation from the null hypothesis but does not quantify the practical strength or real-world importance of an association. Real applications spanning genetics, survey research, and experimental design illustrate how these inference procedures answer important research questions about categorical phenomena, while careful attention to conditions ensures that conclusions remain valid and appropriately qualified within proper context.